I am considering storing data-urls in my mongoDB instead of storing a reference to a file or using GridFS.
Data url:
data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAMcAAAEsCAYAAAB38aczAAAgAElEQV
All of the files I am storing are JPG or PNG, and are less than 1MB in size.
I am wondering if this is considered bad practice, and what the performance implications for both read and write operations, storing the data-urls 1) in a separate collection 2) as meta data in a collection.
I'm open to any other suggestions for small file storage.
First, I wouldn't store base64 encoded data in a database that is perfectly capable of storing binary data. That's just waste of space. Store the image itself, not its base64 representation, i.e. not data : "VBORw0KGgoAAAANSUhEUgAAA...", but data : BinData("VBORw0KGgoAAAANSUhEUgAAA...") (the former is a string for MongoDB, the latter is binary data). Base64 increases the size by 33%.
Other than that, I think this is fine. The trade off is 1 request that grabs all the data vs. multiple requests. The downside of storing larger chunks of data is that all the data must be in RAM for a moment, but at 1MB that's probably a non-issue.
You should, however, make sure that you don't fetch the document in situations where you don't need the image. 1MB isn't too much, but for a read-heavy collection it's a disaster.